Hi Ronny,
How are you getting the crash dump if your system is hung
and you are not getting into kmdb? To get into kmdb,
you need to type f1-a on the console (based on the addresses,
you are on x64, not SPARC).
max
Ronny Egner wrote:
HI all,
cpuinfo -v yields (see file "cpuinfo.txt" ).
I noticed one interesting thing:
When preparing to catche the hang i opened the console and pre-typed "mdb -K"
to crash the system if needed to.
When the system hang i pressed <ENTER> but nothing happened. While digging around in the
core dump i found my "mdb -K":
ffffff1381a42e40 ffffff13a6594cc0 ffffff138192c310 1 60 0
PC: _resume_from_idle+0xf1 CMD: mdb -K
stack pointer for thread ffffff1381a42e40: ffffff008c643d30
[ ffffff008c643d30 _resume_from_idle+0xf1() ]
swtch+0x145()
cv_wait+0x61()
vmem_xalloc+0x635()
vmem_alloc+0x161()
segkmem_xalloc+0x90()
segkmem_alloc_vn+0xcd()
segkmem_zio_alloc+0x24()
vmem_xalloc+0x546()
vmem_alloc+0x161()
kmem_slab_create+0x81()
kmem_slab_alloc+0x5b()
kmem_cache_alloc+0x1fa()
zio_data_buf_alloc+0x2c()
arc_get_data_buf+0x18b()
arc_buf_alloc+0xa2()
arc_read_nolock+0x12f()
arc_read+0x75()
dbuf_read_impl+0x172()
dbuf_read+0xfe()
dmu_buf_hold_array_by_dnode+0x1c9()
dmu_buf_hold_array+0x6e()
dmu_read_uio+0x4d()
zfs_read+0x2d1()
fop_read+0x6b()
vn_rdwr+0x17f()
gexec+0x140()
exec_common+0x45c()
exece+0x1f()
_sys_sysenter_post_swapgs+0x149()
ffffff1381a42e40::thread
ADDR STATE FLG PFLG SFLG PRI EPRI PIL INTR
ffffff1381a42e40 run 1000 104 3 60 0 0 n/a
ffffff1381a42e40::threadlist
ADDR PROC LWP CMD/LWPID
ffffff1381a42e40 ffffff13a6594cc0 ffffff138192c310 mdb/1
So i looked for the mdb process and found it on CPU ID #7:
I noticed "mdb -K" was in run queue on CPU ID 7:
ID ADDR FLG NRUN BSPL PRI RNRN KRNRN SWITCH THREAD PROC
7 ffffff1376a44540 1f 7 0 99 no no t-0 ffffff008b137c60 sched
| |
RUNNING <--+ +--> PRI THREAD PROC
READY 60 ffffff1381a42e40 mdb
QUIESCED 60 ffffff13a35ee720 nfsd
EXISTS 60 ffffff1377474400 bash
ENABLE 60 ffffff008c1bec60 sched
59 ffffff13817ef540 nscd
59 ffffff13817fbc60 syslogd
58 ffffff1381a38720 smbd
So it seems mdb was blocked by sched (thread ffffff008b137c60); digging it
yields:
ffffff008b137c60::findstack
stack pointer for thread ffffff008b137c60: ffffff008b1370d0
ffffff008b137120 intr_thread_prolog+0x2a()
ffffff008b137140 apic_setspl+0x5c()
ffffff008b137180 splr+0x55()
ffffff008b137c60 0x22d9fd9301c7()
Any ideas?
Message was edited by: ronnyegn
------------------------------------------------------------------------
_______________________________________________
opensolaris-discuss mailing list
[email protected]
_______________________________________________
opensolaris-discuss mailing list
[email protected]
